UvA-DARE is a service provided by the library of the University of Amsterdam (http://dare.uva.nl)

UvA-DARE (Digital Academic Repository)

Modelling runoff and erosion for a semi-arid catchment using a multi-scale approach basedon hydrological connectivity

Lesschen, J.P.; Schoorl, J.M.; Cammeraat, L.H.

Published in:Geomorphology

DOI:10.1016/j.geomorph.2009.02.030

Link to publication

Citation for published version (APA):Lesschen, J. P., Schoorl, J. M., & Cammeraat, L. H. (2009). Modelling runoff and erosion for a semi-aridcatchment using a multi-scale approach based on hydrological connectivity. Geomorphology, 109(3-4), 174-183.https://doi.org/10.1016/j.geomorph.2009.02.030

General rightsIt is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s),other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).

Disclaimer/Complaints regulationsIf you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, statingyour reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Askthe Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam,The Netherlands. You will be contacted as soon as possible.

Download date: 16 Jun 2020

Geomorphology 109 (2009) 174–183

Contents lists available at ScienceDirect

Geomorphology

j ourna l homepage: www.e lsev ie r.com/ locate /geomorph

Modelling runoff and erosion for a semi-arid catchment using a multi-scale approachbased on hydrological connectivity

J.P. Lesschen a,⁎, J.M. Schoorl b, L.H. Cammeraat a

a Institute for Biodiversity and Ecosystem Dynamics, University of Amsterdam, Nieuwe Achtergracht 166, 1018WV, Amsterdam, The Netherlandsb Land Dynamics Group, Wageningen University and Research Centre, P.O. Box 47, 6700 AA Wageningen, The Netherlands

⁎ Corresponding author. Present address: Alterra,Research Centre. P.O. Box 47, 6700 AA, Wageningen, T484687; fax: +31 317 419000.

E-mail addresses: janpeter.lesschen@wur.nl (J.P. Less(J.M. Schoorl), l.h.cammeraat@uva.nl (L.H. Cammeraat).

0169-555X/$ – see front matter © 2009 Elsevier B.V. Adoi:10.1016/j.geomorph.2009.02.030

a b s t r a c t

a r t i c l e i n f o

Article history:Received 19 September 2008Received in revised form 8 February 2009Accepted 27 February 2009Available online 12 March 2009

Keywords:Agricultural terraceHydrological connectivityScale dependencySediment dynamicsSoutheast SpainVegetation pattern

Runoff and erosion processes are often non-linear and scale dependent, which complicate runoff and erosionmodelling at the catchment scale. One of the reasons for scale dependency is the influence of sinks, i.e. areasof infiltration and sedimentation, which lower hydrological connectivity and decrease the area-specificrunoff and sediment yield. The objective of our study was to model runoff and erosion for a semi-aridcatchment using a multi-scale approach based on hydrological connectivity. We simulated runoff andsediment dynamics at the catchment scale with the LAPSUS model and included plot and hillslope scalefeatures that influenced hydrological connectivity. The semi-arid Carcavo catchment in Southeast Spain wasselected as the study area, where vegetation patches and agricultural terraces are the relevant sinks at theplot and hillslope scales, respectively. We elaborated the infiltration module to integrate these runoff sinks,by adapting the parameters runoff threshold and runoff coefficient, which were derived from a rainfallsimulation database. The results showed that the spatial distribution of vegetation patches and agriculturalterraces largely determined hydrological connectivity at the catchment scale. Runoff and sediment yield forthe scenario without agricultural terraces were, respectively, a factor four and nine higher compared to thecurrent situation. Distributed hydrological and erosion models should therefore take account of relevantsinks at finer scales in order to correctly simulate runoff and erosion-sedimentation patterns.

© 2009 Elsevier B.V. All rights reserved.

1. Introduction

Scale dependency in erosion research has often been addressed asan important issue that deserves further attention (e.g. Poesen et al.,2003; Boardman, 2006). Simple extrapolations of plot measurementsgenerally lead to large overestimations of runoff and erosion rates atcatchment level, because of the non-linearity of runoff and erosionprocesses. Several studies demonstrated scale dependency with fieldexperiments at different scales (e.g. Le Bissonnais et al., 1998;Cammeraat, 2002; Wilcox et al., 2003; Cerdan et al., 2004; Yair andRaz-Yassif, 2004). These studies showed a strong decrease in runoffand erosion rates with increasing plot length. Also the threshold forthe occurrence of runoff increases at higher scale levels (Cammeraat,2004). The overestimation of runoff and erosion rates can beattributed to the spatially varying influences of sinks, i.e. areas ofinfiltration and sedimentation. For example, plot scale studies ofrunoff and erosion do not reveal sediment deposition at footslopes orincreased infiltration on agricultural terraces. However, runoff and

Wageningen University andhe Netherlands. Tel.: +31 317

chen), jeroen.schoorl@wur.nl

ll rights reserved.

erosionmodelsmostly focus on one specific spatial scale level, i.e. fieldscale or catchment scale (Jetten et al., 1999), without addressing theinfluence of relevant sinks at other scales. Although a few conceptualmodels have been developed that address this scale dependency inerosion simulation (Parsons et al., 2004; Lu et al., 2005), no erosionmodelling study has actually addressed this scale dependency for realcatchments.

It is well known that area-specific runoff yield decreases withincreasing area. For large watersheds this decrease is attributed tofactors, such as rain cell size, lateral changes in lithology and channelwidth, and increasing storage possibilities in the valley domain. Forsmall areas the non-uniform infiltration and spatial variability of thevegetation and surface properties are responsible for the decrease inrunoff rates (Yair and Raz-Yassif, 2004; Kirkby et al., 2005). Besides,Wainwright and Parsons (2002) and Reaney et al. (2007) demon-strated that temporal variations in rainfall intensity during a stormevent can also lead to scale dependency in runoff. However, forerosion this relationship is less clear. Parsons et al. (2006) experi-mentally demonstrated that sediment yield decreases with plotlengths N7 m, due to the limited travel distance of individualentrained particles and due to the decline in runoff coefficient asplot length increases. However, at broader scales additional erosionprocesses, such as gully erosion, mass movement and bank erosion,can become active and increase the area-specific sediment yield.

175J.P. Lesschen et al. / Geomorphology 109 (2009) 174–183

Nevertheless, from a certain basin area threshold, sediment yieldbecomes dominated by sediment transport and sediment depositionrather than by active erosion processes, and will consequentlydecrease with increasing basin area (De Vente and Poesen, 2005).

The concept of hydrological connectivity is becoming increasinglyapplied within the field of hydrology and geomorphology (Hooke,2006; Bracken and Croke, 2007). Hydrological connectivity can bedefined as the physical linkage of water and sediment through thefluvial system (Hooke, 2003). This definition is scale independent,which makes the concept of hydrological connectivity useful to takeaccount of scale dependency in soil erosion research. Since hydro-logical connectivity describes the linkage of runoff and sediment, itfinally determines whether runoff and sediment will becomeconnected at broader scales. Identification of the areas that functionas a sink is therefore crucial to model runoff and erosion at thecatchment scale. The connectivity of runoff and sediment are notalways linked, such as on resistant substrates such as limestone, butalso biological soil crusts can lead to high runoff rates but low erosionrates (Belnap, 2006). Although the term hydrological connectivity isalso used for subsurface flow (Buttle et al., 2004; Ocampo et al., 2006),we only consider Hortonian overland flow in this paper, which is themain runoff generating mechanism in semi-arid environments (Bryanand Yair, 1982).

The main factors that, given a certain rainfall event, influencehydrological connectivity at the plot scale are micro-topography andvegetation. Vegetation in semi-arid ecosystems is characterised by aheterogeneous pattern of bare and vegetated patches, which makeoverland flow highly discontinuous owing to the non-uniforminfiltration (Cerdà, 1998; Ludwig et al., 2005). Vegetation is also oneof the main factors influencing connectivity at the hillslope scale. Thespatial structure of these vegetation patterns finally determines thehydrological connectivity on the hillslope (Puigdefabregas et al.,1999). Results from experimental plots showed that coarsening ofvegetation patterns lead to significantly higher erosion rates due to

Fig. 1. Topography of the upper part of the Carcavo basin (coordinate values are in UTM Europthe location of the study area.

concentration of overland flow and higher flow velocities (Puigde-fabregas, 2005; Bautista et al., 2007). When runoff becomesconcentrated downslope it can lead to gully formation and thesegullies are effective links for transferring water and sediment fromhillslopes to valley bottoms and channels, and consequently increasehydrological connectivity (Poesen et al., 2003). Also human-madestructures, such as ditches, agricultural terraces, paths and roads,influence hydrological connectivity (Croke and Mockler, 2001). At thecatchment scale the geomorphology of a landscape and humanimpacts in the floodplain (e.g. check dams and reservoirs) determinehydrological connectivity. Sediment conveyance is especially impededby fluvial barriers, buffers and discontinuities between landscapecompartments (Fryirs et al., 2007).

The objective of our study was to model runoff and erosion for asemi-arid catchment using a multi-scale approach based on hydro-logical connectivity. Runoff and erosion have been simulated at thecatchment scale, but plot and hillslope scale features that influencethe hydrological connectivity were taken into account. First, weidentified the relevant sinks of runoff and sediment at plot andhillslope scales and quantified their influence. Afterwards weincorporated the effects of these sinks in the LAPSUS runoff anderosion model by adjusting infiltration capacity. Finally, runoff anderosion were simulated for four different scenarios to evaluate theinfluence of the identified sinks on hydrological connectivity at thecatchment scale.

2. Regional setting

We selected the Carcavo basin in Southeast Spain as our study areabecause of its representativeness of semi-arid catchments that aresusceptible to erosion. The catchment is located in the Province ofMurcia, near the town of Cieza (38°13′ N; 1°31′ W). It is a smallcatchment of 30 km2with altitudes ranging from220–850m. The areahas an average annual rainfall of 300 mm and a potential

ean_1950 datum zone 30 N). The inset map of Spain indicates the Region of Murcia and

176 J.P. Lesschen et al. / Geomorphology 109 (2009) 174–183

evapotranspiration of 900 mm. Rainfall is bimodal with peaks in Apriland October and torrential rainstorms especially occur in autumn. Thegeology and geomorphology of the area consists of steep Jurassiccarbonate mountains with calcareous pediments, hilly terrain withKeuper gypsummarls and Cretaceous and Miocene marls in the lowerparts of the basin. Soils in the Carcavo basin are closely linked to thegeology, with thin soils (Leptosols) on the limestone and dolomitemountains, weakly developed soils on marls (Regosols), other soils inwhich often (petro)calcic horizons have developed (Calcisols) on thestable pediments, and gypsiferious soils (Gypsisols) on the Keupermarls. Current land use in the study area consists of agricultural land(barley, olives, almonds and vineyards), abandoned land, reforestedland and shrubland. In the 1970s large parts of the area were plantedwith pine (Pinus halepensis) for reforestation and soil conservationpurposes. During recent decades part of the non-irrigated agriculturehas been abandoned and is currently under different stages ofsecondary succession. For the simulation of runoff and erosion weselected the upper part of the Carcavo catchment, which comprised498 ha (Fig. 1). This part of the basin is characterised by the steepslopes and footslopes of the Sierra del Oro, which are partly reforested,and by a less dissected part with agricultural and abandoned land.

We focused on the twomost relevant sinks for runoff and sedimentwithin the catchment, which are vegetation patches at the plot scaleand agricultural terraces at the hillslope scale. Vegetation in semi-aridareas is characterised by a heterogeneous pattern of bare soil andvegetated patches, which influences hydrological connectivity anddetermines whether runoff is generated at the plot scale (Ludwig etal., 2005). At broader scales, the spatial configuration of vegetationdetermines whether runoff generated at the plot scale will becomeconnected and reach the channel (Puigdefabregas, 2005). However, inthe agricultural part of the catchment, the agricultural terraces havemajor impacts on hydrological connectivity, since well functioningterraces can be large sinks for runoff and sediment. These agriculturalterraces and earth dams are important features of traditionalagricultural systems in hilly areas in Mediterranean countries. In theCarcavo basin ~39% of agricultural land is currently terraced. Terraceshave been constructed on hillslopes and in dry streambeds, whileearth dams are mainly found in valley bottoms of undulating cerealfields.

3. Methodology

3.1. LAPSUS model

To simulate runoff and sediment dynamics we used the LAPSUSmodel (Schoorl and Veldkamp, 2001; Schoorl et al., 2002). LAPSUS is a

Fig. 2. Rainfall intensity (vertical bars) and Measured soil moisture under bare so

dynamic landscape evolution model that can simulate erosion andsedimentation, based on a limited number of input parameters. Themodel is based on two fundamental assumptions: (1) the potentialenergy of overland flow is the driving force for sediment transport(Kirkby, 1986), and (2) the difference between sediment input andoutput of a grid cell is equal to the net increase in storage, i.e. thecontinuity equation for sediment movement (Foster andMeyer,1975).The model evaluates the rate of sediment transport by calculating thetransport capacity of water flowing downslope from one grid cell toanother as a function of discharge and slope gradient. A surplus ofcapacity is compensated by sediment detachment, depending on theerodibility of the surface, which provokes erosion. When the rate ofsediment in transport exceeds the local capacity, the surplus isdeposited based on a sedimentability factor. The routing of runoff andresulting model calculations are based on a multiple flow algorithm,which allows a better representation of divergent and convergentproperties of topography (Holmgren,1994). Themodel can be used forboth short and long-term applications by varying the time step. Forour study we simulated runoff and erosion based on 1-h time steps fora 6 day rainfall period in November 2006.

The main input data for the LAPSUS model are a digital elevationmodel (DEM), rainfall data, infiltration data, erodibility characteristicsand soil depth. To represent the topography we used a 5 m resolutionDEM, which was derived from a 1:5000 topographic map with 5 mcontour lines. Although this resolution is relatively high, it still did notsufficiently represent the agricultural terraces, since these generallyhave height difference of only 1–2 m. Therefore we had to adapt theDEM to include the topography of the terraces. First, we digitised theterraces in ArcGIS 9.0 (ESRI, Redlands, US) and calculated the meanaltitude for each terrace using zonal statistics. The DEM was adjustedaccording to the difference between the mean terrace altitude and theoriginal topography, which was multiplied by a factor 0.75 to includesome influence of the original topography, because most terraces arenot completely flat, especially at the sides.

For rainfall we used data from a large event in November 2006,which was recorded with a tipping bucket rain gauge in the centre ofthe catchment. The event included several rainstorms within 6 dayswith a total rainfall of 157 mm, which was half of the average annualrainfall. Although most rainfall was of low intensity (Fig. 2), therainstorm of the last day had a higher intensity with a maximum of50 mm h−1, based on a 10-min interval. The recurrence time of thislast rainstorm was about 2 years, based on a 36-year rainfall series ofthe Almadenes weather station. However, when rainfall of theprevious days was taken into account, the recurrence time for the10-day rainfall sumwas ~23 years. Although the spatial distribution ofrainfall can be highly variable (González-Hidalgo et al., 2001), we

il and vegetation on marl during a six-day rainfall period in November 2006.

Table 1Occurrence of substrateswithin the simulated catchment and the estimated parameters.

Substrate Occurrence(%)

Erodibility factor(m−1)

Soil depth(m)

Slope deposits 27.4 0.0025 0.40Slope deposits with calcrete 13.6 0.0015 0.25Marl 39.4 0.0050 0.40Keuper 9.3 0.0035 0.20Limestone/dolomite 10.3 0.0005 0.15

Fig. 3. Land use of the simulated catchment and locations of intact and failed terracewalls on (abandoned) agricultural land.

177J.P. Lesschen et al. / Geomorphology 109 (2009) 174–183

considered the catchment to be sufficiently small to neglect the spatialvariability of rainfall.

The most important input for our modelling exercise is infiltration.Modification of the infiltration capacity enables the incorporation ofrunoff sinks, which finally determine whether runoff will occur. In theoriginal version of the LAPSUS model a fixed value for infiltration isused for each soil type and land use (Schoorl et al., 2002). Buis andVeldkamp (2008) modified the infiltration parameters to betterrepresent the spatial diversity in water redistribution and availabilitywithin arid catchments. We further elaborated the infiltration moduleby making it dynamic and dependent on measured parameters. Therunoff threshold and coefficient were selected to describe theinfiltration characteristics in the LAPSUS model. The runoff thresholdwas defined as the amount of rainfall until runoff starts, and the runoffcoefficient is the ratio of observed runoff to applied rainfall. Theseparameters were derived from a database of rainfall simulations,which is described in Section 3.2. For each time-step the model firstcalculates the amount of rainfall available for runoff after subtractingthe runoff threshold. This value is then multiplied by the runoffcoefficient, which determines the amount of runoff for each grid cell.Next, the model determines the routing of runoff, based on themultiple flow algorithm, and evaluates whether infiltration capacity isstill available. When runoff has been determined for the entirecatchment, the model calculates the sediment transport capacity,which determines whether sediment is eroded or deposited. Weestimated that 15% of the runoff threshold and 10% of the storage andinfiltration capacity of the terraces is recovered per day due toevapotranspiration and drainage to deeper soil horizons.

The erodibility factor and soil depth were estimated for eachsubstrate (Table 1). The distribution of the substrates was derivedfrom a 1:50,000 geological map and an additional field survey, whichwas required in order to map the location of calcretes. The erodibilityfactor determines how easy sediment is detached with low valuesindicating low erodibility. For all agricultural land we added a value of0.002 to the erodibility factor to take account of ploughing, which ingeneral increases erodibility. The last input variable is a sediment-ability or settlement factor, which determines how easy sediment canbe deposited (Schoorl et al., 2002). For this factor we distinguishedbetween bare soil and vegetation, since vegetation increases surfaceroughness and promotes sedimentation (Bochet et al., 2000). Thesedimentability factor was set a factor 5 higher for vegetated areas,which means that transport capacity is lowered and sediment is moreeasily deposited.

3.2. Influence of vegetation patches

Vegetation in semi-arid environments is characterised by hetero-geneous patterns of bare and vegetated patches. The bare patches aregenerally formed by bare rock and crusted soils with poor soilstructure and low infiltration rates, whereas vegetated patches havebetter soil properties with higher organic matter content and astronger aggregation, which results in a higher infiltration capacity.This makes overland flow highly discontinuous with bare patches asrunoff generating areas and vegetated patches as runoff sinks(Bergkamp, 1998; Cammeraat and Imeson, 1999). Soil moisture

measurements in the study area confirmed the higher infiltrationcapacity under vegetation (Fig. 2). FDR (Frequency Domain Reflecto-metry) sensors at 5 cm depth recorded soil moisture at a bare land anda vegetated patch. Soil moisture under vegetation was on average7 vol.% higher, and reacted faster to rainfall, which indicates higherinfiltration capacity (Bergkamp, 1998).

To simulate the influence of vegetation patterns on runoff anderosion, we differentiated the infiltration parameters for bare andvegetated areas. These parameters, runoff threshold and runoffcoefficient, were derived from a database of rainfall simulationexperiments. The database comprised data from studies that wereapplied in similar semi-arid areas, mostly in Southeast Spain (Boix-Fayos et al., 1995; Quinton et al., 1997; Cerdà, 1997a,b, 1998, 2001;Imeson et al., 1998; Cammeraat and Imeson,1999; Lasanta et al., 2000;Cammeraat et al., 2002; Ceballos et al., 2002; Calvo-Cases et al., 2003;Castro et al., 2006; Seeger, 2007; Meerkerk et al., 2008). We onlyselected data from experiments that were applied on substrates andland uses that occur within the study area (Fig. 3), which resulted in285 observations. Although these experiments were used to study awide range of research questions and not all studies measured ordescribed the same variables, the size of the database is sufficientlylarge to obtain general and statistically significant relationshipsbetween environmental factors and measured parameters. For eachobservation from this databasewe calculated the runoff threshold andcoefficient, which together determine the amount of infiltration in theLAPSUS model. However, combining both parameters will over-estimate total infiltration capacity, since part of the rainfall wasalready included in the runoff threshold. Hence, the runoff coefficientwas recalculated based on the amount of rainfall minus the runoffthreshold. Next, we classified all observations for the differentcombinations of land use (shrubland, abandoned land and agriculturalland), substrate (marl, slope deposits and limestone) and vegetationcover (bare and vegetated). For the 11 relevant combinations wecalculated the mean runoff threshold (n=143) and runoff coefficient(n=218), which were used in the LAPSUS model (Fig. 4).

Upscaling of the vegetation patterns was required to include theinfluence of vegetation patches for the entire catchment. Imeson andPrinsen (2004) and Lesschen et al. (2008a) found a linear relationshipbetween vegetation cover and spatial metrics that described thevegetation patterns. This enabled us to use vegetation cover as a proxyfor spatial vegetation structures in landscapes with spotted vegetation

Fig. 4. Mean runoff threshold and runoff coefficient (RC) for the different combinations of land use, substrate and vegetation cover (error bars indicate standard error of the mean).Aban: abandoned land; slope dep: slope deposits; veg: vegetated.

178 J.P. Lesschen et al. / Geomorphology 109 (2009) 174–183

patterns. Lesschen (2008) created a fractional vegetation cover mapfor the Carcavo basin based on a linear regression of vegetation cover,as derived from detailed aerial photographs, and the green and redreflectances from a QuickBird image (R2 of 0.91|Pb0.001, n=20).Since the rainfall simulation database contained insufficient data onfractional vegetation cover, we could only differentiate two classes ofvegetation cover, i.e. bare and vegetated areas. The threshold betweenthese two classes was set at 30%, since several studies showed thatrunoff and erosion increase drastically b30% vegetation cover (Francisand Thornes, 1990; Quinton et al., 1997; Ludwig et al., 2002).

3.3. Influence of agricultural terraces

Agricultural terraces are important sinks for runoff and erosion atthe hillslope scale. These terraces were originally constructed toreduce soil erosion and to intercept runoff by decreasing the generalslope (Morgan, 1995). However, the effects of terracing are not alwayspositive, since the area near the terrace rim is now under influence of asteep hydraulic gradient, which may lead to gully erosion and piping(Faulkner et al., 2003). To include the influence of the terraces inLAPSUS we first digitised the terrace walls in ArcGIS, based on aQuickBird satellite image of 2006. During a field survey we checkedwhich terraces were intact and which had collapsed (Fig. 3). In total~500 terrace walls were identified, of which 127 had collapsed. Mostfailed terraces were located on abandoned fields, which are morevulnerable to failure due to the lack of maintenance and increasedrunoff (Lesschen et al., 2008b). The terrace wall map was converted toraster and intact terraces were assigned an additional infiltrationcapacity. Although in reality this additional storage and infiltrationcapacity depends on the terrace characteristics (e.g. size and slope ofthe terrace and height of the terrace rim) no such data were availableto create a variable storage capacity for the terraces. Instead we used asingle value of 50 m3 of water that can be additionally stored andinfiltrated at each terrace, which is based on the size of an averageterrace and terrace rim.

3.4. Scenarios

To evaluate the influence of the identified sinks on hydrologicalconnectivity at the catchment scale, we used four different scenariosfor the simulation of runoff and erosion. Besides a scenario thatrepresents the current situation, two other scenarios were used toassess the influence of vegetation and agricultural terraces, respec-tively, and a fourth scenario was used to simulate the influence of a

soil and water conservation practise in which all failed terraces arerepaired. Accordingly the four scenarios are:

1. Current situation2. No vegetation patterns3. No terraces4. All terraces function

The first scenario includes the influence of vegetation patterns atthe plot scale as well as the influence of intact and collapsed terracesat the hillslope scale. For the second scenario, we adapted the values ofthe runoff threshold and runoff coefficient to exclude the influence ofvegetation patterns. Based on the rainfall simulation database, wecalculated new values for each combination of land use and substratewithout differentiating between bare and vegetated areas. For thethird scenario, we did not include the extra infiltration capacity forterraces and used the original DEM without the detailed terracerepresentation. Finally, for the last scenario we modified the input filewith additional infiltration capacity and gave the failed terraces thesame infiltration capacity as the other terraces.

3.5. Verification of the model results

To verify the spatial distribution of the model results we comparedthe simulated runoff map with observed concentrated flow paths.These connectivity patterns were mapped after the rainfall event ofNovember 2006 (Meerkerk, 2009). Since the soilwas already saturatedfrom the rainfall of the previous days, the intense rainfall on 8November resulted in significant amounts of erosion, as observed bycompletely filled sediment collectors, rills on agricultural fields, andfailure of some agricultural terraces.Within the agricultural part of thesub-catchment all signs of concentrated flow were recorded with ahandheld GPS. These flow paths were visible as rills and gullies, butalso as areaswith clear signs of concentrated overlandflowas observedby the removal of small branches, leaves and other litter material.Although concentrated flow was also observed on roads, we excludedthese flow paths since the influence of roads is not included in themodel. In addition to themapped connectivity patterns on agriculturalland, we added the flow paths in the channels.

4. Results

Based on the analysis of rainfall simulation data we calculated themean runoff threshold and runoff coefficient for the different landuse–substrate–vegetation cover classes (Fig. 4). The runoff threshold

Table 2Summary of model outputs for the simulated scenarios.

Scenario Total runoff(1000 m3)

RC Erosion(Mg ha−1)

Sedimentation(Mg ha−1)

Sediment yield(Mg ha−1)

SDR

1 52.9 0.068 38.1 35.7 2.5 0.0632 76.5 0.098 56.4 49.1 7.4 0.1293 203.0 0.259 84.8 61.7 23.4 0.2724 40.4 0.052 32.0 29.1 3.0 0.091

179J.P. Lesschen et al. / Geomorphology 109 (2009) 174–183

for bare conditions was lower compared to vegetated conditions. Forbare areas runoff starts after ~7mmof rainfall, while under vegetationthe runoff threshold is ~13 mm. The results also showed that thethreshold on abandoned fields is much lower than for shrubland andagricultural land. The runoff coefficient also clearly showed thedifference between bare and vegetated areas, on average the runoffcoefficient was higher by 0.19 under bare conditions. The runoff

Fig. 5. Cumulative runoff f

coefficient was also higher for abandoned land compared to shrublandfor both vegetated and bare conditions.

The results of the runoff and erosion simulations with the LAPSUSmodel for the four scenarios are summarised in Table 2. Total runoff isthe sum of discharge at the outlet of the catchment. RC is the runoffcoefficient for the entire catchment, i.e. total runoff divided by thetotal rainfall. Erosion and sedimentation are the sum of erosion andsedimentation, respectively, for each time step. Sediment yield is thesum of the transported sediment at the catchment outlet and isconsequently the net erosion for the entire catchment. Finally, SDR isthe sediment delivery ratio, which is the fraction of erosion that istransported out of the catchment.

For the current situation (Scenario 1) a runoff coefficient of 0.068and a sediment yield of 2.5 Mg ha−1 were predicted. The sedimentdelivery ratio was only 0.063, which means that most sediment isdeposited within the catchment. For the second scenario (without theinfluence of vegetation patterns) more runoff was produced and the

or the four scenarios.

180 J.P. Lesschen et al. / Geomorphology 109 (2009) 174–183

sediment yield was a factor 3 higher compared to the currentsituation. In the third scenario (without agricultural terraces) therunoff coefficient and sediment yield weremuch higher, while erosionand sedimentation remained relatively low, which means that moresediment was transported out of the catchment, as indicated by thehigh sediment delivery ratio. In the last scenario (with all terracesintact) less runoff was produced and erosion and sedimentation werelower. However, sediment yield and sediment delivery ratio weresomewhat higher than in the scenario for the current situation.

Fig. 5 shows the results of the cumulative runoff for the fourscenarios. The cumulative runoff is the sum of runoff for all time stepsthat passed a grid cell and is expressed in meter height. In the firstscenario, representing the current situation, runoff from the Sierra delOro was well connected to the main channel, while runoff from theagricultural fields was mainly retained by the agricultural terraces.Only fields close to the channel were connected and contributed to

Fig. 6. Erosion (negative values) and sedimentation

channel discharge. For the second scenario (without the influence ofvegetation patterns), runoff intensity was higher and especially theslopes of the Sierra del Oro produced more runoff. For the southernpart of the catchment somewhat more runoff was produced, but thisdid not increase hydrological connectivity. The third scenario (with-out the influence of agricultural terraces) showed a very strongincrease in runoff in the southern part of the catchment where runoffof all hillslopes became connected to the channel. Runoff from thehillslopes in the last scenario (with all terraces intact) was notconnected to the channel in the agricultural part of the catchment. Alllocally produced runoff was retained by terraces downslope and thechannel was only supplied by runoff from nearby areas.

The results of the simulation of erosion and sedimentation resultedin different magnitudes of sediment dynamics, and the patterns oferosion and sedimentation were also different for the four scenarios(Fig. 6). In the current situation most erosion occurred on the steep

(positive values) patterns for the four scenarios.

Fig. 7. Overlay of observed concentrated flow (red) based on Meerkerk et al. (submittedfor publication) and simulated runoff (blue) for scenario 1 (current situation). (Forinterpretation of the references to colour in this figure legend, the reader is referred tothe web version of this article.)

181J.P. Lesschen et al. / Geomorphology 109 (2009) 174–183

slopes of the Sierra del Oro in the northern part of the catchment andon areas near the channel, while sedimentation occurred on thefootslopes and on agricultural terraces. In the channel an alternationof areas with erosion and those with sedimentationwas observed. Forthe second scenario (without the influence of vegetation patterns) theerosion and sedimentation patterns were similar, but the amount oferosion and sedimentation was higher. The third scenario (withoutany terrace influences) resulted in a completely different pattern inthe agricultural part of the catchment. High erosion rates occurred inall valley bottoms, while sedimentation was limited to the footslopesof the Sierra del Oro and the channels. Finally, the fourth scenarioshowed the positive effects of well-maintained terraces, whichretained sediment and prevented erosion in the agricultural part ofthe catchment.

5. Discussion

The results of our runoff and erosion simulations showed thatvegetation patterns and agricultural terraces havemajor influences onthe connectivity of runoff and sediment. According to the first andsecond scenarios, excluding the influence of vegetation patterns led toan overestimation of runoff and erosion, especially in areas with ahigher vegetation cover. However, the patterns of hydrologicalconnectivity at catchment scale did not change. In contrast, thescenario without the influence of agricultural terraces showed majorimpacts on hydrological connectivity at the catchment scale. For thisscenario, total runoff was four times higher and sediment yield almost10 times higher, due to lowering of the runoff threshold onagricultural fields (Cammeraat, 2004). Consequently all runoff fromthe hillslopes of the entire catchment was connected to the channel.Nevertheless the last scenario showed the important function ofagricultural terraces for soil and water conservation. In this case mostof the runoff and sediment from the agricultural part of the catchmentwas not connected to the channel and retained by the agriculturalterraces. Although total runoff and gross erosion were lower, thesediment yield for this scenario was somewhat higher. This might beexplained by a lower sediment delivery from agricultural land to thechannel, which can increase the erosive force of the water in thechannel, and lead to more erosion in the channel and a highersediment transport from the catchment. This effect is similar to the‘clear water effect’ behind checkdams (Castillo et al., 2007).

Although erosion on agricultural land is generally higher comparedto semi-natural areas (Kosmas et al., 1997; Cammeraat, 2004),hydrological connectivity was lower, due to the influence ofagricultural terraces and other water conservation structures. As aresult the sediment yield from agricultural areas is limited to fieldsnear the channel, and most sediment is retained on agricultural fields.However, during extreme events the accumulated sediment mightstill be transported to the channel due to the failure of terraces, whichwill increase hydrological connectivity. Furthermore, in case ofagricultural abandonment, part of the terraces might collapse, whichwill also increase hydrological connectivity. This was observed inseveral parts of the catchment, where gullies through terrace wallsincreased the connectivity.

To verify the spatial distribution of the simulation, we comparedthe modelled runoff map of scenario 1 with the observed concen-trated flow (Fig. 7). This map showed that the overall runoff patternwas fairly well simulatedwith low connectivity between the hillslopesand channel in the southern part of the catchment. Field observationsconfirmed that flow inside the channel was fully connected, assimulated by the model. Furthermore, the highest cumulative runoffclass (N100 m) coincided very well with the location of the actualchannels. However, not all individual flow paths were correctlysimulated, probably due to lack of detailed input data. A more detailedDEM, for instance one derived by LIDAR, is necessary to capture alltopographical features such as small ditches, agricultural terraces and

field boundaries, which do determine local surface runoff. Also roadsare important pathways for runoff, which frequently lead to gullyinitiation due to runoff concentration (Croke and Mockler, 2001;Meerkerk, 2009). Furthermore, identification of concentrated flow inthe field is not always straightforward, since not all concentrated flowis visible as rills. On freshly ploughed fields the development of rills isenhanced due to higher erodibility, because soils crusts are brokendown. However, on abandoned land the formation of rills is reducedby the presence of (biological) soil crusts and most overland flow willnot leave clear traces. This makes identification of flow paths in thefield more difficult and might lead to mismatches between simulatedand observed connectivity patterns.

Besides the verification of spatial runoff patterns, the simulatedtotal runoff and sediment yield at the catchment outlet should beassessed, since these determine the impact on off-site effects such asflash flooding risk and reservoir sedimentation. Unfortunately, nomonitoring data were available yet to validate total discharge andsediment yield for the modelled catchment. However, based on otherstudies for comparable semi-arid catchments, we were able to assesswhether the simulated runoff and sediment yield were realistic. Thesimulated runoff coefficient for the entire catchment was 0.068, basedon the current situation scenario. Although this value is quite low, itseems a reasonable value considering the low intensity of most of therainfall and the presence of agricultural terraces in the southern partof the catchment, which are major runoff sinks. Other studies alsofound rather low runoff coefficients for semi-arid catchments. Martín-Vide et al. (1999) reported runoff coefficients ranging between 0.04–0.15 for a 30 km2 catchment in Northeast Spain, and Michaud andSorooshian (1994) and Ye et al. (1997) found runoff coefficients of~0.1 for the semi-arid Walnut Gulch catchment in New Mexico andephemeral catchments in Western Australia, respectively. For sedi-ment yield less comparable studies were available. De Vente andPoesen (2005) demonstrated the relation between drainage area andarea-specific sediment yield based on Spanish data, which first showsan increase in area-specific sediment yield and afterwards a decreasefor drainage areas N100 km2. However, no data were available forcatchments with drainage areas between 0.1–30 km2, which have thehighest area-specific sediment yields. Nevertheless, a maximum area-specific sediment yield of 20–40 Mg ha−1 year−1 for small scalecatchments (1–10 km2) can be inferred from their graphs. This is

182 J.P. Lesschen et al. / Geomorphology 109 (2009) 174–183

higher than our simulated sediment yield of 2.5 Mg ha−1, but asdiscussed before, rainfall intensity was not very high for most of theevent, while in semi-arid landscapes most erosion can be attributed toextreme events (Cammeraat, 2004; Boardman, 2006).

Modelling runoff and erosion at the catchment scale makessimplifications of hydrological processes and sediment dynamicsunavoidable. Runoff generation, which is the most important processfor ourmodelling approach, was simplified by using a runoff thresholdand runoff coefficient, while runoff generation in reality is controlledby complex and dynamic interactions between rainfall characteristicsand soil and surface properties (Martinez-Mena et al., 1998). Moresophisticated approaches, such as the Green–Ampt equation, requiretoo much input data at the catchment scale. The values for the runoffthreshold and runoff coefficient were mean values from all rainfallsimulation experiments, for which we did not make a distinctionbetween dry and wet preceding conditions. However, Cammeraat andImeson (1999) and Cammeraat (2004) showed that runoff curves candiffer greatly depending on the initial soil moisture conditions.Nevertheless, in our modelling approach, infiltration capacity wascalculated dynamically and rainfall from the previous time steps wastaken into account. Another simplification was the non-variableinfiltration and storage capacity for the agricultural terraces, which ledto underestimation of runoff on small terraces, while large terraceswith well constructed rims could retainmore runoff. Nevertheless, theoverall runoff pattern appeared realistic, since field observationsshowed that only limited areas were connected to the channel andmost agricultural hillslopes retained all runoff. When a more detailedDEM is available, the sink function of agricultural terraces can also besimulated dynamically based on the algorithm of Temme et al. (2006),which models the sediment buffer function of depressions. A finalsimplification is the exclusion of bench terraces from the reforestedareas in the simulation, which might have led to an overestimation ofrunoff and erosion in the northern part of the catchment. However,Maestre and Cortina (2004) and Hooke (2006) have questioned theeffectiveness of these terraces, and De Wit and Brouwer (1998) foundhigher runoff and sediment yields in reforested areas due to gullyingthrough the bench terraces.

Modelling runoff and erosion without taking into account therelevant sinks from the plot and hillslope scale will lead to over-estimation of runoff and sediment yield, as demonstrated in our study.Otherwise,models can predict acceptable soil loss and discharge at theoutlet, butwith incorrect patterns of source and sink areas (Jetten et al.,2003). Application of erosion models for soil and water conservationpurposes requires correct simulation of runoff and erosion patterns inorder to identify the hotspots where mitigation measures should beapplied. Although Jetten et al. (2003)were negative about the ability ofcatchment based models to determine optimal locations for waterconservation measures (e.g. grass strips or terraces) our simulationsgave satisfactory results for the implementation of a conservationpractise, i.e. restoration of existing failed terraces. In the currentsituation (scenario 1) several areas with high runoff production in theagricultural part can be identified, and restoration of the failed terracesat these sites (scenario 4) significantly reduced runoff and erosion.Also other measures could be evaluated (e.g. grass strips) by adaptinginfiltration and erodibility input data.

6. Conclusions

Our new approach is based on the identification of hydrologicalsinks which are integrated in a distributed runoff–erosion model. Theresults of the simulations show that the spatial distribution of sinks atplot and hillslope scales, in this case vegetation patches andagricultural terraces, determines much of the hydrological connectiv-ity at the catchment scale. Runoff and sediment dynamics aretherefore non-linear and scale dependent, and are to a large extentdetermined by the spatial distribution of hydrological sinks. Especially

the additional infiltration and storage capacity of agricultural terracesshould be taken into account, since their exclusion will highlyoverestimate runoff and erosion in catchments with these terraces,and erroneous runoff and erosion patternswill be simulated. AlthoughLAPSUS is a relatively simple model, we have demonstrated that themodel is able to reproduce the observed patterns of hydrologicalconnectivity and to predict realistic values for total runoff andsediment yield. Distributed hydrological and erosion models shouldtake account of relevant sinks at finer spatial scales in order tosimulate patterns of runoff and erosion correctly.

Acknowledgements

This research, within the RECONDES Project, was funded by theEuropean Commission, Directorate-General of Research, Environmentand Sustainable Development Programme, Natural Resources Man-agement and Services, Project no. GOCE-CT-2003-505361 (1 February2004–30 April 2007). Bas van Wesemael and André Meerkerk arethanked for the use of their observed connectivity map.

References

Bautista, S., Mayor, Á., Bourakhouadar, J., Bellot, J., 2007. Plant spatial pattern predictshillslope runoff and erosion in a semiarid Mediterranean landscape. Ecosystems 10,987–998.

Belnap, J., 2006. The potential roles of biological soil crusts in dryland hydrologic cycles.Hydrol. Process. 20, 3159–3178.

Bergkamp, G., 1998. A hierarchical view of the interactions of runoff and infiltrationwith vegetation and microtopography in semiarid shrublands. Catena 33, 201–220.

Boardman, J., 2006. Soil erosion science: reflections on the limitations of currentapproaches. Catena 68, 73–86.

Bochet, E., Poesen, J., Rubio, J.L., 2000. Mound development as an interaction ofindividual plants with soil, water erosion and sedimentation processes on slopes.Earth Surf. Process. Landf. 25, 847–867.

Boix-Fayos, C., Calvo, A., Imeson, A.C., Schoorl, J.M., Soto, S., Tiemessen, I.R., 1995.Properties and erosional response of soils in a degraded ecosystem in Crete(Greece). Environ. Monit. Assess. 37, 79–92.

Bracken, L.J., Croke, J., 2007. The concept of hydrological connectivity and itscontribution to understanding runoff-dominated geomorphic systems. Hydrol.Process. 21, 1749–1763.

Bryan, R.B., Yair, A., 1982. Badland Geomorphology and Piping. Geobooks, Norwich, UK.Buis, E., Veldkamp, A., 2008. Modelling dynamic water redistribution patterns in arid

catchments in the Negev Desert of Israel. Earth Surf. Process. Landf. 33, 107–122.Buttle, J.M., Dillon, P.J., Eerkes, G.R., 2004. Hydrologic coupling of slopes, riparian zones

and streams: an example from the Canadian Shield. J. Hydrol. 287, 161–177.Calvo-Cases, A., Boix-Fayos, C., Imeson, A.C., 2003. Runoff generation, sediment

movement and soil water behaviour on calcareous (limestone) slopes of someMediterranean environments in southeast Spain. Geomorphology 50, 269–291.

Cammeraat, L.H., 2002. A review of two strongly contrasting geomorphological systemswithin the context of scale. Earth Surf. Process. Landf. 27, 1201–1222.

Cammeraat, L.H., 2004. Scaledependent thresholds inhydrological anderosion responseofa semi-arid catchment in southeast Spain. Agric. Ecosyst. Environ. 104, 317–332.

Cammeraat, L.H., Imeson, A.C., 1999. The evolution and significance of soil-vegetationpatterns following land abandonment and fire in Spain. Catena 37, 107–127.

Cammeraat, L.H., Willott, S.J., Compton, S.G., Incoll, L.D., 2002. The effects of ants' nestson the physical, chemical and hydrological properties of a rangeland soil in semi-arid Spain. Geoderma 105, 1–20.

Castillo, V.M., Mosch, W.M., Conesa García, C., Barberá, G., Navarro Cano, J.A., López-Bermúdez, F., 2007. Effectiveness and geomorphological impacts of check dams forsoil erosion control in a semiarid Mediterranean catchment: El Cárcavo (Murcia,Spain). Catena 70, 416–427.

Castro, G., Romero, P., Gomez, J.A., Fereres, E., 2006. Rainfall redistribution beneath anolive orchard. Agric. Water Manage. 86, 249–258.

Ceballos, A.,Martinez-Fernandez, J., Santos, F., Alonso, P., 2002. Soil-water behaviourof sandysoils under semi-arid conditions in theDueroBasin (Spain). J. Arid Environ. 51, 501–519.

Cerdà, A., 1997a. Soil erosion after land abandonment in a semiarid environment ofSoutheastern Spain. Arid Land Res. Rehab. 11, 163–176.

Cerdà, A., 1997b. The effect of patchy distribution of Stipa tenacissima L on runoff anderosion. J. Arid Environ. 36, 37–51.

Cerdà, A., 1998. The influence of geomorphological position and vegetation cover on theerosional and hydrological processes on a Mediterranean hillslope. Hydrol. Process.12, 661–671.

Cerdà, A., 2001. Effects of rock fragment cover on soil infiltration, interrill runoff anderosion. Eur. J. Soil Sci. 52, 59–68.

Cerdan, O., Le Bissonnais, Y., Govers, G., Lecomte, V., van Oost, K., Couturier, A., King, C.,Dubreuil, N., 2004. Scale effect on runoff from experimental plots to catchments inagricultural areas in Normandy. J. Hydrol. 299, 4–14.

Croke, J., Mockler, S., 2001. Gully initiation and road-to-stream linkage in a forestedcatchment, southeastern Australia. Earth Surf. Process. Landf. 26, 205–217.

183J.P. Lesschen et al. / Geomorphology 109 (2009) 174–183

De Vente, J., Poesen, J., 2005. Predicting soil erosion and sediment yield at the basinscale: scale issues and semi-quantitative models. Earth-Sci. Rev. 71, 95–125.

De Wit, A.M.W., Brouwer, L.C., 1998. The effect of afforestation as a restoration measurein a degraded area in a Mediterranean environment near Lorca (Spain). In: Uso, J.L.,Brebbia, C.A., Power, H. (Eds.), Ecosystems and Sustainable Development. Advancesin Ecological Sciences, vol. 1. Computational Mechanics Publications, Southampton,UK, pp. 165–170.

Faulkner, H., Ruiz, J., Zukowskyj, P., Downward, S., 2003. Erosion risk associated withrapid and extensive agricultural clearances on dispersive materials in southeastSpain. Environ. Sci. Policy 6, 115–127.

Foster, G.R., Meyer, L.D., 1975. Mathematical simulation of upland erosion by fundamentalerosion mechanics. In: Anonymous (Ed.), Present and Perspective Technology forPredicting Sediment Yields and Sources. Proceedings Sediment Yield Workshop,Oxford, 1972. Department of Agriculture, Washington D.C., pp. 190–207.

Francis, C.F., Thornes, J.B., 1990. Runoff hydrographs from three Mediterraneanvegetation cover types. In: Thornes, J.B. (Ed.), Vegetation and Erosion. John Wiley& Sons Ltd, Chichester.

Fryirs, K.A., Brierley, G.J., Preston, N.J., Kasai, M., 2007. Buffers, barriers and blankets: the(dis)connectivity of catchment-scale sediment cascades. Catena 70, 49–67.

González-Hidalgo, J.C., De Luis, M., Raventós, J., Sánchez, J.R., 2001. Spatial distributionof seasonal rainfall trends in a western Mediterranean area. Int. J. Climatol. 21,843–860.

Holmgren, P., 1994. Multiple flow direction algorithms for runoff modelling in gridbased elevation models: an empirical evaluation. Hydrol. Process. 8, 327–334.

Hooke, J., 2003. Coarse sediment connectivity in river channel systems: a conceptualframework and methodology. Geomorphology 56, 79–94.

Hooke, J.M., 2006. Human impacts on fluvial systems in the Mediterranean region.Geomorphology 79, 311–335.

Imeson, A.C., Prinsen, H.A.M., 2004. Vegetation patterns as biological indicators foridentifying runoff and sediment source and sink areas for semi-arid landscapes inSpain. Agric. Ecosyst. Environ. 104, 333–342.

Imeson, A.C., Lavee, H., Calvo, A., Cerda, A., 1998. The erosional response of calcareoussoils along a climatological gradient in Southeast Spain. Geomorphology 24, 3–16.

Jetten, V., de Roo, A., Favis-Mortlock, D., 1999. Evaluation of field-scale and catchment-scale soil erosion models. Catena 37, 521–541.

Jetten, V., Govers, G., Hessel, R., 2003. Erosion models: quality of spatial predictions.Hydrol. Process. 17, 887–900.

Kirkby, M.J., 1986. A two-dimensional simulation model for slope and stream evolution.In: Abrahams, A.D. (Ed.), Hillslope Processes. Allen & Unwin Inc., Winchester, US,pp. 203–222.

Kirkby, M.J., Bracken, L.J., Shannon, J., 2005. The influence of rainfall distribution andmorphological factors on runoff delivery from dryland catchments in SE Spain.Catena 62, 136–156.

Kosmas, C., Danalatos, N., Cammeraat, L.H., Chabart, M., Diamantopoulos, J., Farand, R.,Gutierrez, L., Jacob, A., Marques, H., Martinez-Fernandez, J., Mizara, A., Moustakas,N., Nicolau, J.M., Oliveros, C., Pinna, G., Puddy, R., Puigdefabregas, J., Roxo, M., Simao,A., Stamou, G., Tomasi, N., Usai, D., Vacca, A., 1997. The effect of land use on runoffand soil erosion rates under Mediterranean conditions. Catena 29, 45–59.

Lasanta, T., Garcia-Ruiz, J.M., Perez-Rontome, C., Sancho-Marcen, C., 2000. Runoff andsediment yield in a semi-arid environment: the effect of land management afterfarmland abandonment. Catena 38, 265–278.

Le Bissonnais, Y., Benkhadra, H., Chaplot, V., Fox, D., King, D., Daroussin, J., 1998.Crusting, runoff and sheet erosion on silty loamy soils at various scales andupscaling from m2 to small catchments. Soil Till. Res. 46, 69–80.

Lesschen, J.P., Cammeraat, L.H., Kooijman, A.M., VanWesemael, B., 2008a. Developmentof spatial heterogeneity in vegetation and soil properties after land abandonment ina semi-arid ecosystem. J. Arid Environ. 72, 2082–2092.

Lesschen, J.P., Cammeraat, L.H., Nieman, T., 2008b. Erosion and terrace failure due toagricultural land abandonment in a semi-arid environment. Earth Surf. Process.Landf. 33, 1574–1584.

Lesschen, J.P., 2008. Multi-scale interactions between soil, vegetation and erosion in thecontext of agricultural land abandonment in a semi-arid environment. Ph.D. Thesis.Univ. of Amsterdam, The Netherlands. (http://dare.uva.nl/en/record/282600).

Lu, H., Moran, C.J., Sivapalan, M., 2005. A theoretical exploration of catchment-scalesediment delivery. Water Resour. Res. 41, W09415.

Ludwig, J.A., Eager, R.W., Bastin, G.N., Chewings, V.H., Liedloff, A.C., 2002. A leakinessindex for assessing landscape function using remote sensing. Landsc. Ecol. 17,157–171.

Ludwig, J.A., Wilcox, B.P., Breshears, D.D., Tongway, D.J., Imeson, A.C., 2005. Vegetationpatches and runoff-erosion as interacting ecohydrological processes in semiaridlandscapes. Ecology 86, 288–297.

Maestre, F.T., Cortina, J., 2004. Are Pinus halepensis plantations useful as a restorationtool in semiarid Mediterranean areas? For. Ecol. Manag. 198, 303–317.

Martín-Vide, J.P., Ninerola, D., Bateman, A., Navarro, A., Velasco, E., 1999. Runoff andsediment transport in a torrential ephemeral stream of the Mediterranean coast. J.Hydrol. 225, 118–129.

Martinez-Mena, M., Albaladejo, J., Castillo, V.M., 1998. Factors influencing surface runoffgeneration in a Mediterranean semi-arid environment: Chicamo watershed, SESpain. Hydrol. Process. 12, 741–754.

Meerkerk, A.L., Van Wesemael, B., Cammeraat, L.H., 2008. Water availability in almondorchards on marl soils in southeast Spain: the role of evaporation and runoff. J. AridEnviron. 72, 2168–2178.

Meerkerk, A.L., 2009. Rainfed orchards in semi-arid environments - Retaining the waterand the soil. Ph.D. Thesis. VDM Verlag Dr. Müller, Saarbrücken, Germany.

Michaud, J., Sorooshian, S., 1994. Comparison of simple versus complex distributedrunoff models on a midsized semiarid watershed. Water Resour. Res. 30, 593–605.

Morgan, R.P.C., 1995. Soil Erosion and Conservation, (2nd Ed.). Longman Group Limited,Essex, UK.

Ocampo, C.J., Sivapalan, M., Oldham, C., 2006. Hydrological connectivity of upland-riparian zones in agricultural catchments: implications for runoff generation andnitrate transport. J. Hydrol. 331, 643–658.

Parsons, A.J., Wainwright, J., Powell, D.M., Kaduk, J., Brazier, R.E., 2004. A conceptualmodelfor determining soil erosion by water. Earth Surf. Process. Landf. 29, 1293–1302.

Parsons, A.J., Brazier, R.E., Wainwright, J., Powell, D.M., 2006. Scale relationships inhillslope runoff and erosion. Earth Surf. Process. Landf. 31, 1384–1393.

Poesen, J., Nachtergaele, J., Verstraeten, G., Valentin, C., 2003. Gully erosion andenvironmental change: importance and research needs. Catena 50, 91–133.

Puigdefabregas, J., 2005. The role of vegetation patterns in structuring runoff andsediment fluxes in drylands. Earth Surf. Process. Landf. 30, 133–147.

Puigdefabregas, J., Sole, A., Gutierrez, L., del Barrio, G., Boer, M., 1999. Scales andprocesses of water and sediment redistribution in drylands: results from theRambla Honda field site in Southeast Spain. Earth-Sci. Rev. 48, 39–70.

Quinton, J.N., Edwards, G.M., Morgan, R.P.C., 1997. The influence of vegetation speciesand plant properties on runoff and soil erosion: results from a rainfall simulationstudy in south east Spain. Soil Use Manage. 13, 143–148.

Reaney, S.M., Bracken, L.J., Kirkby, M.J., 2007. Use of the Connectivity of Runoff Model(CRUM) to investigate the influence of storm characteristics on runoff generationand connectivity in semi-arid areas. Hydrol. Process. 21, 894–906.

Schoorl, J.M., Veldkamp, A., 2001. Linking land use and landscape process modelling: acase study for the Alora region (south Spain). Agric. Ecosyst. Environ. 85, 281–292.

Schoorl, J.M., Veldkamp, A., Bouma, J., 2002. Modeling water and soil redistribution in adynamic landscape context. Soil Sci. Soc. Am. J. 66, 1610–1619.

Seeger, M., 2007. Uncertainty of factors determining runoff and erosion processes asquantified by rainfall simulations. Catena 71, 56–67.

Temme, A.J.A.M., Schoorl, J.M., Veldkamp, A., 2006. Algorithm for dealing withdepressions in dynamic landscape evolution models. Comput. Geosci. 32, 452–461.

Wainwright, J., Parsons, A.J., 2002. The effect of temporal variations in rainfall on scaledependency in runoff coefficients. Water Resour. Res. 38, 1271.

Wilcox, B.P., Breshears, D.D., Allen, C.D., 2003. Ecohydrology of a resource-conservingsemiarid woodland: effects of scale and disturbance. Ecol. Monogr. 73, 223–239.

Yair, A., Raz-Yassif, N., 2004. Hydrological processes in a small arid catchment: scaleeffects of rainfall and slope length. Geomorphology 61, 55–169.

Ye,W., Bates, B., Viney, N.R., Sivapalan,M., Jakeman, A.J.,1997. Performance of conceptualrainfall-runoff models in low-yielding ephemeral catchments. Water Resour. Res.33, 153–166.